Lightnews — Scholar-powered news

Zach Danziger @danzigerzachary.bsky.social · 14d

It's available on all the usual podcasting platforms, so subscribe and check it out.

Hopefully it will be interesting for the science curious and science skeptical, alike.

Holy Shift: Biomedical Breakthroughs Shaping Tomorrow - Episode 1: YOUR NEW HEART | Mike Davis

Zach Danziger @danzigerzachary.bsky.social · 14d

Our new podcast is out about why biomedical science is so important to our communities:
Holy Shift! Biomedical Breakthroughs Shaping Tomorrow

Please Share!

Our first episode is Dr. Mike Davis on repairing congenital heart defects in children. ❣️

www.youtube.com/watch?v=xrvO...

YouTube video by Wallace H. Coulter Dept. of Biomedical Engineering

www.youtube.com

1 1 1

Reposted by Zach Danziger

Chris Rozell @crozell.bsky.social · Jul 17

Please repost! We're hiring a postdoc to work with me and
@salagapan.bsky.social on an exciting interdisciplinary project in human neuroscience and brain-body interactions underlying effortful behavior in health and mood disorders.

siplab.gatech.edu/postdoc_ad_2...

SIPLab

siplab.gatech.edu

1 29 26

Zach Danziger @danzigerzachary.bsky.social · Jul 13

Congratulations. You're going to rock it there. 🫀🥼

Zach Danziger @danzigerzachary.bsky.social · Jul 11

Its a risk, yeah. But given its stability for so many years and contracts with government and big corporations you'd at least see something like that coming a long way away.
C++ is a better language in general to be sure, ... but not for neuroscience.
Hot take: MATLAB will last longer than Python.

Zach Danziger @danzigerzachary.bsky.social · Jul 11

Arguments covered in the post:
1. Python is free, just like science.
2. Python is opens source.
4. People don’t switch to Python just because they’re lazy and don’t want to port legacy code.
5. Python does deep learning though, which is hot.
?. Why do you hate Python, you hater?

Zach Danziger @danzigerzachary.bsky.social · Jul 11

That sucks. Very much feel your pain on that.
But you stack up the number of times that happens with MATLAB (once in my entire science career) vs the number of times that stuff happens with my Python stack... well.

Confessions of a Neuro MATLAB Apologist – Danziger Lab

Zach Danziger @danzigerzachary.bsky.social · Jul 10

If you're doing #neuroscience out there, you don't have to be peer pressured to switch from #matlab to python. You do you. ❤️
scholarblogs.emory.edu/danziger/202...

Embrace the Fusion with Neural Technology – Danziger Lab

2 1 4

Zach Danziger @danzigerzachary.bsky.social · Jun 30

An argument for embracing the coming wave of neurotechnology: Neural augmentation can enhance and amplify our humanity without threatening our identity. After all, it worked for the Borg.
scholarblogs.emory.edu/danziger/202...

Zach Danziger @danzigerzachary.bsky.social · Jun 6

This is amazing.

Zach Danziger @danzigerzachary.bsky.social · Jun 5

Embrace the philosophy :)

Zach Danziger @danzigerzachary.bsky.social · Jun 5

I'm convinced by the thread.

It feels like you're saying that neural activity being correlated with, or even partially caused by, some stimulus isn't sufficient for it to count as a representation.

Representation is teleological on this account. Seems appropriate.

Zach Danziger @danzigerzachary.bsky.social · May 15

Hang in there - you do great science.

I'm not sure if this is good advice or not, but I usually use bourbon in these situations.

Zach Danziger @danzigerzachary.bsky.social · May 15

I was very pleased with myself after doing the Question and Hypothesis boxes and getting 8/10. I was considerably less smug after instructing an LLM to do the next two boxes and earning a 6 and 7 / 10.

Zach Danziger @danzigerzachary.bsky.social · May 15

This is a great idea.

Now, how to convince trainees to use it...

I've never written the project abstract in the planning phase before. I'll give that a try going forward. Writing grants forces you to do a lot of this work.

Zach Danziger @danzigerzachary.bsky.social · May 6

Maybe some day....

Zach Danziger @danzigerzachary.bsky.social · Apr 15

Reminds me of SINDy for nonlinear dynamical systems discovery.

Zach Danziger @danzigerzachary.bsky.social · Apr 11

Would be a disaster if this is a widespread rollback.

1 2

Zach Danziger @danzigerzachary.bsky.social · Mar 26

2) I hope you're right about large models facilitating discovery, and it is a definite possibility. I don't think its predestined though. I can imagine ways where huge models only modestly better experiments, or it becomes very hard to map those experiments back from the model to the biology.

2

Zach Danziger @danzigerzachary.bsky.social · Mar 26

1) (sorry, lack of chars) There are so many ways one could decide to quantify, and so many behaviors to quantify (in potentially different ways), and no competition nearly as well defined as the folding competition rules, so many brain regions from which to quantify, handling subject differences...

Making Sense of Softmax – Danziger Lab

Zach Danziger @danzigerzachary.bsky.social · Mar 26

"What ... prevent us from pursuing an AlphaFold for the brain?"

Well-defined goals.

Even movement decoding isn't nearly as tightly quantifiable as 3d conformation, and that would be just one part of the brain. And even if AI crushes decoding it may not help in "understanding brain function".

1 4

Zach Danziger @danzigerzachary.bsky.social · Mar 24

The output Softmax should be interpreted as "the amount of class-N-iness" the input has, not the input's "probability of belonging to class N".

scholarblogs.emory.edu/danziger/202...